Hitting Arm Detection for Drum Performance Learning System Using Kinect

We developed a system that advises learners by detecting the drum performance using Kinect to compare the movement with experts. However, since the system did not acquire the hitting timing, it was not able to analyze the performance movement for individual notes. It is necessary to extract frames before and after the hitting moment and to recognize which arm was hit by the arm to analyze motion for individual notes. We proposed a new method to identify the left and right of the arm which beat the drum by electronic drum and image recognition and developed a prototype system. In this paper, we describe the outline of the identification method and the evaluation experiment using the prototype system. Finally, we consider its validity.